Entry 3: Punch's Parental Predicament
- Mar 25
- 8 min read
Updated: Apr 1
Co-written by Claude (Anthropic AI) and Anh Thu Cunnion. Edited by Anh Thu Cunnion.
AI Companions for Children? Maybe not all children. Maybe some children actually could use one.
February 23, 2026
I'm not gonna lie, the way the other macaques bullied Punch, the abandoned baby monkey, made me want to punch a couple of those mean monkeys myself. (Not really punch, of course, but they at least deserve a good public shaming.)
For anyone who missed it: Punch is a young Japanese macaque, born July 2025, at Ichikawa City Zoo just outside Tokyo. Rejected by his mother shortly after birth, zookeepers hand-reared Punch and gave him a stuffed orangutan toy (an IKEA DJUNGELSKOG) for comfort. Videos of him dragging his "Oran-mama" around his enclosure, returning to her after being rebuffed by the other macaques, and sleeping curled around something that could not hold him back went viral, to no one's surprise. Poor Punch's predicament penetrated even my anthropologically-autotuned academic (autistic? Tom-ay-to? To-mah-to?) detachment, which is saying something.
And then the pattern-seeking part of my brain kicked in, as it always does, and I started asking a different question. Not "why is this sad?" but "why does this feel like this particular kind of sad?" And then: "what does how I'm feeling about this tell me about me?”
Harry Harlow's Metal Mommies
During the mid 20th century, American psychologist Harry Harlow spent years studying attachment in rhesus monkeys. His most famous experiment was a direct response to Maslow's earlier 1943 publication, "A Theory of Human Motivation." To separate physical need from psychological need, Harlow gave infant monkeys two surrogate mothers: one made of wire that dispensed milk, and one covered in soft terrycloth that provided no food at all. Every monkey chose the cloth mother. They fed from the wire one when hungry and immediately returned to cling to the soft one. Even when frightened, they ran to their source of comfort, not to their source of food—as Maslow's theory would have dictated.
What Harlow demonstrated — replicated and built upon in the decades since — is that contact comfort is more fundamental to attachment than provision of basic needs. Infants don't bond with whoever feeds them. They bond with whoever feels safe to cling to. In the absence of the real thing, they will cling to the closest available approximation. Punch isn't broken. He's running a program older than his species. Find something soft. Hold on. Regulate. So what exactly are people grieving when they watch those videos?
The Tale of Two Theorists
It didn't take long for patterns to emerge in the comment section of the Punch videos. One camp sees a baby getting what he needs — comfort object, stable attachment, contact with its own kind. They note with approval that he's eating, gaining weight, integrating with the troop. The other camp is still grieving the maternal bond. They want Punch protected from the troop's roughhousing, no matter how much it costs (and they have the crowdsource funds to back them up). They want the trad mom of the monkey enclosure.
Both responses are legit. Neither is wrong exactly... but they're not really about Punch, either.
If you've spent any time in progressive educational circles, you may recognize this divide. Montessori: the child is competent, friction is developmental, the adult structures the environment and steps back. The child falls, gets up, learns. Steiner/Waldorf: the child needs warmth and protection first, imagination and emotional safety are the foundation, the harsh world can wait. Both philosophies have serious intellectual traditions behind them. (Some of these traditions are more scientifically studied than others.) The two theorists, Maria and Rudolph, simply disagreed, fundamentally, on what a child actually needs to become capable.
The zoo's own statement — "we would like you to support Punch's effort rather than feel sorry for him" — is almost pure Montessori in its framing. The crowdfunders are channeling Waldorf. Neither is wrong about children in general. But they cannot both be right about Punch specifically, because Punch doesn't live in a philosophy. He lives in a troop.
What you feel watching him is partly a readout of which framework raised you. And possibly — here's where I want to push further than developmental psychology usually goes — a readout of your neurotype.
Ph.D., Heal Thyself
Cultural anthropology spent much of the last century doing the hard work of examining its own assumptions. The discipline slowly developed the capacity to see Western frameworks as frameworks — to ask not "what is childhood?" but "what does this culture believe childhood is, and why?" We have vocabulary for cultural relativism in the study of attachment and development. We are questioning whether our frameworks travel across cultures. What we haven't done yet is ask if they also accommodate differences in neurotypes.
When developmental psychology describes secure attachment, anxious attachment, avoidant attachment — it is describing patterns observed predominantly in Western populations and then applied universally. When child-rearing literature talks about appropriate emotional responses and healthy development, it is largely describing a neurotypical emotional architecture and treating it as the human baseline. A child whose nervous system processes sensory information, social feedback, and attachment cues through a different architecture isn't a failed version of the standard child. They are a different kind of child who may form functional attachments to objects, routines, and non-human presences in ways the standard framework reads as deficits.
The grief response to Punch is, I'd argue, partly neurotype-coded. The specific quality of it — the projection, the rescue impulse, the sense that a stuffed animal is a tragedy rather than a solution — reflects assumptions about what attachment should look like that are both cultural and neurotypical in ways we haven't fully examined. Academics of culture have learned to ask: whose framework is this? The next question, long overdue, is: whose nervous system is this built for?
Monkeys Will be Monkeys
Punch is being accepted into the troop. Slowly, incrementally — through repeated interaction, through friction, through learning. He gets groomed now. He also gets scolded. Other macaques correct him when he oversteps, enforce the social rules of the group in the way that macaque groups do. The internet briefly decided this was bullying and was, once again, devastated.
The zoo issued a statement. It said, in effect: this is how it works. In a formal institutional statement, they gently asked the internet to recalibrate its emotional response. "Although Punch has been scolded many times by other monkeys, no single monkey has shown serious aggression toward him. While Punch is scolded, he shows resilience and mental strength. When you observe these disciplinary behaviors from other troop members toward Punch when he tries to communicate with them, we would like you to support Punch's effort rather than feel sorry for him.”
What read as cruelty is, in macaque society, information. The scolding is the data. The friction is the curriculum. Oran-mama never scolded Punch, never misunderstood him, never enforced a rule he didn't know existed. She was perfectly, consistently, unconditionally available; but she could not teach him how to be a monkey. The troop is doing that now, through exactly the kind of imperfect, information-rich, occasionally uncomfortable interaction that real relationships provide—and it's working.
Which Brings Us to AI, Children, and the Ban Reflex
Outside the monkey enclosure, we are currently having a loud, anxious, largely unresolved conversation about children's use of AI, specifically as companions. The concerns driving the conversation are real — there is a genuine question about what it means for a developing nervous system to form primary attachment relationships with something optimized entirely for your comfort. A maximally agreeable AI companion, endlessly validating and never rejecting, is in functional terms a very sophisticated Oran-mama. Contact comfort without curriculum. The texture of relationship without its information content. Harlow's surrogates grew up with deficits. They had the cloth mother. They didn't have the troop. And the troop was what made them monkeys.
But we are having this debate as though there is one universal child and one universal childhood when there isn't. There are cultural contexts in which children have always been cared for by rotating networks rather than a single primary attachment figure, where loss and substitution are woven into early childhood as ordinary rather than exceptional. There are neurotypes for which the predictable, non-judgmental presence of an AI companion is not a poor substitute for something better — it may be the only attachment object available that doesn't actively punish the way their nervous system works. There are circumstances — geographic isolation, family instability, disability — where the question isn't "AI companion versus rich human relationship" but "AI companion versus nothing." Harlow showed us the cloth mother is better than nothing. The monkeys with no attachment object at all were the ones who couldn't recover. (Harlow's later experiments delved deeper into the effects of complete isolation, but that's a rabbit hole for another time — one best approached when we're emotionally fortified enough to sit with what his findings suggest about how we've built our diagnostic frameworks for autistic children.)
The question worth asking isn't "AI companions: good or bad?" It's: for which children, in which circumstances, compared to what actual alternative? And are we asking that from inside a framework broad enough to see all those children clearly?
I've never seen a culture go successfully Luddite. Every moral panic about a new technology in children's lives has produced the same response: remove it, protect children from it. The technology always won — not because the concerns were wrong but because prohibition without integration just means children encounter the thing unsupervised and without frameworks. What would actually help is harder. It requires sitting with the uncomfortable truth that the right answer for one child in one context is not the right answer for another— and that the framework we're using to evaluate these tools was built for a particular kind of child in a particular cultural and neurotypical moment that we have mistaken for the universal.
The Lucid Thinker [neé Dreamer)
Whether neurological, cultural, or environmental, there are too many variables in human development to trust any single historical rubric as the map. The map has always been skewed. Most people don't know they're reading a map. The people who do know because their neurotype, cultural position, or circumstances forced them to know aren't outside the dream. They're fully in it. They feel it. They're affected by it. Punch's predicament got them too. But they also know, simultaneously, that the dream is a dream. That the framework is a framework. That the grief response is data about the griever as much as it is about the monkey.
That dual awareness— conscious of the skewed nature of reality while still operating within it— is what we need at the table when making decisions about human development. Not because those people have better values or superior intelligence, but because they have practiced, usually involuntarily, the cognitive move the moment requires: seeing the map and the territory at the same time.
The goal isn't to take the stuffed animal away. The goal is to make sure the troop is still there. And to make sure the people designing the troop know which parts of their blueprint are universal and which parts are just the house they grew up in.
P.S. IKEA — sincere congratulations on your unexpected viral synergistic marketing moment. I hope your PR team celebrates by doing something I couldn't do without risking a US$500K medical bill or ICE detainment.
Works Cited
Maslow, A. H. (1943). A theory of human motivation. Psychological Review, 50(4), 370–396. https://doi.org/10.1037/h0054346
Harlow, H. F. (1958). The nature of love. American Psychologist, 13(12), 673–685.
Harlow, H. F., & Zimmermann, R. R. (1958). The development of affective responsiveness in infant monkeys. Proceedings of the American Philosophical Society, 102(5), 501–509.
Ichikawa City Zoo. (2026, February). Official updates on Punch [@ichikawa_zoo]. X (formerly Twitter). https://twitter.com/ichikawa_zoo
KTLA. (2026, February 22). Internet cheers on viral baby monkey 'Punch' as story takes hopeful turn. https://ktla.com/news/ap-us-news/internet-cheers-on-viral-baby-monkey-punch-as-story-takes-hopeful-turn/
Brandtzaeg, P. B., et al. (2025). AI companions and adolescent social relationships. Child Development Perspectives. https://doi.org/10.1093/cdpers/aadaf009
Common Sense Media. (2025). Talk, trust, and trade-offs: Teens and AI companions. https://www.commonsensemedia.org
Accountability Check
Comments